We performed a diagnostic analysis of AVHRR-NDVI and gridded, temperature data for the deciduous forests of the eastern U.S., calibrating temperature accumulation model with satellite data for 1982--1993. The model predicts interannual variability in onset date based upon year-to-year changes in springtime temperature. RMS error over the period ranges from 6.9 days in the northern portion of the domain to 10.7 days in the south. The analysis revealed a relationship between temperature accumulation and satellite derived onset date (rank correlation = 0.31--0.62). The required temperature accumulation threshold can be expressed as a function of mean temperature (R2 of 0.90) to facilitate predictive analysis of phenological onset, or when remote sensing data are unavailable.